Accelerated research for COVID-19: Methodological ruminations for internet-based research

Sir,

It was interesting to read the articles[1,2] under the theme of accelerated research in the context of COVID-19 (Volume 62, Issue 4) in your esteemed Journal. We remark on a critical aspect of their methodology. As a reader, we were bothered by the representativeness of the research considering their aims. This also allows us to ponder on methodologies for online research which may be the way forward for the ensuing time, especially for student research which is mandatory for various postgraduate courses.

The two articles targeted the general population to evaluate the psychological impact[1] and sleep-related changes[2] due to nationwide COVID-19 lockdown. The former sets out to estimate the prevalence of depression, anxiety, perceived stress, and well-being in (India's) general public, while the latter is a survey of sleep experience, routines, physical activity, and their alterations due to lockdown in an unstated universe of participants (presuming it to be the general public). The latter study[2] did explore the IP addresses of the participants and found 22 out of 1024 responses were from outside India.

Despite both the studies used multilingual tools, yet the representativeness of the sample is circumspect as evident from the sociodemographics of the sample of either study ([61.8% postgraduates and 45.3% doctors];[1][graduates 75.9% and healthcare workers 35.9%][2]). Moreover, no information on the validity of the translated tools is provided. This brings into question the incongruity of the aims and the methods of study.

Understandably, the COVID-19 pandemic and lockdown have imposed significant limitations on various arenas, yet any research needs a rational methodology given the limitations of resources. Ameen and Praharaj[3] have discussed cogently the drawbacks of using forwarded survey links on social networking sites and groups for sampling. Such snowball sampling has its specific disadvantages as it is impossible to determine the sampling error. Foregoing compromise is made when participants are rare, hidden, or hard to find due to their specific traits. Moreover, as we see in these two studies, oversampling a particular network of peers (healthcare workers) is a potential source of bias. Thus, the prevalence rates calculated are incorrect, and the findings may not represent the actual pattern of distribution in the population. Thus, the readers need to keep the limitations in mind to infer the results.

In the time of “physical distancing,” online approach to research may be the most appropriate, yet a sampling strategy has to be well thought of for valid research. Another way out is a propensity scoring adjustment[4] that reweights the web-based sample on certain variables to correspond to a reference sample (which is a separate probability sample without any selection bias problem). These variables are referred to as propensity variables, responded by both the web survey and reference sample participants, and are usually questions about characteristics, attitudes, or behaviors on internet use that are hypothesized to differ between the two populations.

To conclude, internet-based research is fraught with its deficiencies, but a well-planned strategy can help overcome these to a large extent.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.